There have always actually been 3 quite different issues addressed by this procedure - workload distribution, cross-cockpit authority gradient, and visual transition at low altitude. Most comments address only the last of these but the other two are actually just as important.
There are lots of accidents and serious incidents in "relatively good" conditions, i.e. weather reported not very close to limits, and in these the crew are almost held to be the prime culprits. Even if the event isn't catastrophic (e.g. landing at wrong airport in VMC) it's usually career-limiting.
In a current review of over 100 reports from the last 25 years where crew "failure" was the primary cause, 39% of the crews had not considered poor weather likely and found themselves unprepared for what they encountered. In many of these, vertical guidance was available, but just wasn't used by either pilot, apparently in part because the PM had been looking out, seen "something", and not returned to monitoring instruments when the PF also went "outside" (e.g. UPS 1354?).
Hence the NTSB recommendations (in the study of ILS accidents in the USA exactly 40 years ago!) that "flightcrew coordination procedures [should] limit sighting callouts to those visual cues which are associated with the runway environment. Unrequired callouts which can result in the premature abandonment of instrument procedures should be prohibited".
The same study also recommended that "flightcrew coordination procedures [should] insure continuous monitoring of the aircraft's instruments from the OM to landing....... procedures which require a transfer or exchange of visual scanning responsibilities should require that the appropriate crew member announce that he is relinquishing previously assigned duties or responsibilities". Does ANY operator have such callouts?
It's also clear that while much lip service is paid to it, CRM training doesn't solve the problem of cross-cockpit monitoring breakdowns. Conventionally the PM's job is both "assist the PF" AND "obstruct the PF if what he/she is trying to do is wrong". Making the switch is a difficult call, especially for junior F/Os - the PM is betting his/her judgement of the situation against the boss. Odds are the boss is right, but not always and the report falgs up a "monitoring/challenging" failure.
Hence yet another NTSB statement and associated recommendation (in 2000): "the monitored approach method provides for more effective monitoring by the non-flying pilot because captains are more likely to be comfortable offering corrections or challenges to first officers than the reverse situation. ............ Monitored approaches decrease the workload of the flying pilot and increase flight crew interaction, especially when experienced captains monitor and prompt first officers during the execution of approaches."
Final approach for autoland cases are actually a bit of a red herring - the biggest safety problem with these seems to be failure of the PF to initiate a go-around correctly when needed, and confusing "landing" actions with "go-around" actions. (Maybe an issue in recent B777 at DXB?) Again this risk can be reduced if one pilot is set up with go-around as no 1 task.
Cat3b autolands are actually a vanishingly small percentage of the total airline ops, and are not where "crew failures" wreck aircraft - it's always in "better" weather but with fewer facilities in the air or on the ground. Anyway if anyone knows of an accident during a low vis. autoland please send me the reference.
So as far as routine use is concerned, IMHO it's smarter to be prepared for the worst by assuming you'll need an MA, reverting to "conventional" when you've found it's really not needed, than the other way round. Just like you prepare for engine failure on every takeoff......